Enum ZeroCopyStrategy 

pub enum ZeroCopyStrategy {
    MmapDirect,
    StagingUpload,
}

How weight data reaches the GPU for a given device.

Variants

MmapDirect

Integrated / unified-memory GPU (Apple Silicon via Metal; integrated Intel/AMD): the mmap’d region is directly GPU-visible — no host→device copy.

StagingUpload

Discrete GPU: upload the mmap’d region once via a staging buffer.

Implementations

impl ZeroCopyStrategy

pub fn for_device(is_unified_memory: bool) -> Self

Pick the strategy from whether the adapter has unified memory.

pub fn for_wgpu_device_type(device_type: DeviceType) -> Self

Map a wgpu device type to the strategy: integrated GPUs share memory with the host (unified) ⇒ mmap-direct; discrete GPUs ⇒ staging upload. CPU/other adapters are treated as unified (the “device” buffer is host memory).

pub fn requires_host_copy(self) -> bool

Whether a host→device copy is required (false on unified memory).

Trait Implementations

impl Clone for ZeroCopyStrategy

fn clone(&self) -> ZeroCopyStrategy

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ZeroCopyStrategy

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for ZeroCopyStrategy

fn eq(&self, other: &ZeroCopyStrategy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for ZeroCopyStrategy

impl Eq for ZeroCopyStrategy

impl StructuralPartialEq for ZeroCopyStrategy